Risk of serious bacterial infection associated with tumour necrosis factor-alpha inhibitors in children with juvenile idiopathic arthritis

Risk of serious bacterial infection associated with tumour necrosis factor-alpha inhibitors in... Abstract Objectives TNF-α inhibitors (TNFIs) have a black box warning for increased risk of serious infection that was based on evidence from studies of adults. Evidence of the association is lacking for children. We aimed to examine the risk of infection posed by TNFIs compared with DMARDs in children with JIA. Methods We conducted a cohort study using the 2009–13 Truven MarketScan Commercial Claims and Encounters database. Children <16 years old with JIA who initiated monotherapy with TNFIs or DMARDs were identified and followed for occurrence of serious bacterial infection requiring hospitalization. Cox proportional hazard models were used to estimate hazard ratios for infection associated with TNFIs compared with DMARDs, adjusting for potential confounders with high-dimensional propensity scores and time-varying CS use. Results We identified 2013 DMARD initiators and 482 TNFI initiators with a mean follow-up of 255 and 307 days, respectively. We identified 18 and 11 patients with a serious infection in the DMARD and TNFI groups, resulting in crude rates of 1.28 (95% CI 0.76-2.02) and 2.72 (95%CI 1.36-4.86) per 100 person-years, respectively. In adjusted models, TNFIs were associated with an increased risk of serious bacterial infection compared with DMARDs (adjusted hazard ratio 2.72, 95% CI: 1.08, 6.86). Conclusion Use of TNFIs poses a higher risk of serious infection compared with DMARDs in children with JIA. Our analysis confirms the US Food and Drug Administration warning about TNFI-associated infection in children with JIA. biologic therapies, tumour necrosis factor-alpha inhibitors, juvenile idiopathic arthritis, bacterial infection, children, drug safety Rheumatology key messages TNF inhibitors pose higher risk of serious bacterial infection than DMARDs in children with JIA. The most common site for TNF inhibitor-associated infection in JIA is the respiratory tract. The US Food and Drug Administration warning about TNF inhibitor-associated infection is confirmed in children with JIA Introduction TNF-α inhibitors (TNFIs), including mAb (e.g. infliximab and adalimumab) and fusion proteins (etanercept), are biological agents used to treat RA and JIA. TNFIs are highly effective for RA and JIA and have been shown in clinical trials to induce disease remission, improve quality of life and enhance physical functioning [1–3]. The ACR recommends use of TNFIs for patients who have received MTX or LEF for 3–6 months and whose disease activity remains uncontrolled [4, 5]. In addition, earlier use of TNFIs has been shown to improve short-term clinical outcomes in both RA and JIA [6, 7]. As a result, TNFI use has increased over the past decade [8, 9]. Although TNFIs are generally safe, growth in their use has resulted in more reports of adverse events. In particular, infections have been the most frequently reported serious adverse event in adult RA patients [10, 11]. However, mixed findings were observed in evaluations of the association between TNFIs and infection in meta-analyses [12–14] and observational studies [10, 11, 15–19]. Some studies reported increased risk of infection for TNFIs compared with DMARDs [12, 13, 15, 16, 19], whereas others found no elevated risk [10, 11, 14, 17, 18]. Nevertheless, in 2008, the US Food and Drug Administration (FDA) required TNFI manufacturers to include a black box warning in the product label for serious infections leading to hospitalization or death [20]. The warning applies to both adults and children, yet most studies evaluating the TNFI–infection association were conducted in adults with RA. Children are a vulnerable population more likely to experience adverse drug events than adults [21]. In addition, the immaturity of their immune system may put children at a higher risk for infection [22]. Unfortunately, clinical and observational studies involving JIA patients have had limited numbers of participants and short follow-up periods [1, 2, 23, 24]. More definitive evidence is needed for the association between TNFIs and infection in children. The aim of this study was to examine the risk of serious bacterial infection associated with TNFIs in children with JIA. Methods Data sources In this retrospective cohort study, we analysed data from the Truven Health MarketScan Commercial Claims and Encounters database for the period from 1 January 2009 to 31 December 2013. The database contains employer-based health insurance claims for enrollees and their dependants across the USA. Administrative data on patient enrolment; health-care utilization, including hospitalizations and visits to outpatient clinics and emergency departments; medical procedures; costs of services; and pharmacy records are available in the database and for research use [25]. Ethical approval was not required to analyse the data in this study. Population The analytical cohort was developed by first identifying children (age <16 years) with JIA who had at least one prescription for a TNFI or DMARD during the study period. JIA was identified using a previously validated algorithm that included International Classification of Disease, ninth version, Clinical Modification codes 714.xx (RA and JIA), 696.0 (PsA) and 720.xx (AS) [26]. Included children were required to have two or more JIA diagnoses within 1 year or one JIA diagnosis coded by a paediatrician or rheumatologist [26]. The date of the first new prescription for a TNFI or DMARD was defined as the index date and included new users who had to have no previous use of either TNFIs or DMARDs 6 months before the index date. In addition, patients were excluded if they met any of the following criteria during the 6 months before the index date: (i) discontinued health plan enrolment; (ii) history of tuberculosis and/or use of medications for tuberculosis; or (iii) history of cancer, transplantation and/or HIV infection. Exposures Exposures to TNFIs and DMARDs were identified using National Drug Codes and Healthcare Common Procedure Coding System. DMARDs included MTX, HCQ, SSZ and LEF, and TNFIs included etanercept, adalimumab, infliximab, certolizumab and golimumab. Each included patient was followed from the index date to the first occurrence of infection, disenrolment from the health plan, discontinuation of treatment, a switch between or a combined use of a TNFI and DMARD, or the end of the study period (31 December 2013). The observation time was the time of continuous use of TNFI monotherapy or DMARD monotherapy. Discontinuation of medication was defined as a gap >92 days between the end of one prescription’s supply (in days) and the next prescription date. For patients who discontinued their medication, we censored them at the end of the last prescription’s days supply before a gap >92 days. Outcomes The outcome of interest was the first occurrence of a serious bacterial infection, which was defined as an infection requiring hospitalization. We applied a previously validated algorithm that incorporated 27 infection sites using International Classification of Disease, ninth version, Clinical Modification codes at any diagnosis position in the inpatient claims [27]. If an individual was diagnosed as having an infection at two or more sites on the same date, the major site of infection was identified based on the severity of infection, concomitant diagnoses and procedures recorded. Only 14% of cases of serious infections had more than one site of infection. High-dimensional propensity score models We used a high-dimensional propensity score (hdPS) to identify and adjust for a large number of covariates that could confound the association between the exposure and infections [28, 29]. Variables were grouped in dimensions that included outpatient diagnoses, inpatient diagnoses, outpatient procedures, inpatient procedures and outpatient medication use during the 6 months before the index date. Thus, any medical conditions (including previous infection) represented by diagnosis or procedure codes or medication use were examined using the hdPS algorithm. The algorithm calculated and ranked a measure of confounding bias for each variable and selected the 500 top-ranked variables into the propensity score model (see supplementary Table S1, available at Rheumatology Online). Along with the 500 empirical variables, we included demographic variables (age, gender, geographical location, calendar year of medication use and type of health plan on the index date) and health-care utilization variables (the number of visits to outpatient, inpatient and emergency departments during the 6 months before the index date) in the logistic regression model to calculate the probability of receiving TNFIs (i.e. the propensity score). Statistical analysis Demographic information, health-care utilization, co-morbidities and concomitant medication use were compared between the TNFI and DMARD groups using Student’s unpaired t-test, a χ2 test or Fisher’s exact test, as appropriate. Crude rates of serious infection (number of events per 100 person-years) and 95% CIs were computed using the Poisson exact method. Specific infection sites were described. A Kaplan–Meier plot and log-rank test were used to examine the rate of serious infection over time. Cox proportional hazard models were used to estimate the hazard ratios (HRs) for serious infection associated with use of TNFIs compared with DMARDs. The tertile of the propensity score was included in the final Cox model. Analyses stratified by gender were also performed. We also conducted several sensitivity analyses to examine the robustness of our findings, as follows: (i) we adjusted for CS use in a time-varying manner to account for the effect of CSs during follow-up; (ii) we restricted our analysis to children ⩾2 to <16 years of age, which is consistent with the FDA label; (iii) we extended the observation period by 30 and 90 days beyond discontinuation of medication use to capture infections occurring after discontinuation; (iv) we reduced the gap defining treatment discontinuation from 92 to 31 days based on the duration of drug effect [1, 30]; (v) We excluded patients with SLE, SS or psoriasis 6 months before the index date owing to the fact that patients with autoimmune diseases may have different baseline risk for infection; (vi) we also excluded patients with IBD from (v); (vii) we excluded patients who used HCQ or SSZ in the DMARD group because users of these two medications usually had less severe disease; and (viii) we excluded 1% and 5% of patients with extreme hdPS, respectively. All analyses were performed using SAS statistical software version 9.4 (SAS Institute, Cary, NC, USA) and STATA 12 (StataCorp, College Station, TX, USA). The Institutional Review Board determined this study to be non-human subject research. Results We identified 5497 children with JIA who were prescribed either a TNFI or a DMARD during the study period (Fig. 1). After applying the exclusion criteria, the final study cohort consisted of 2495 individuals, including 2013 new DMARD users and 482 new TNFI users. Fig. 1 View largeDownload slide Selection criteria for analytical study cohort Fig. 1 View largeDownload slide Selection criteria for analytical study cohort Baseline characteristics were compared between the TNFI and DMARD groups (Table 1). TNFI initiators were slightly older than DMARD initiators (mean age: 10.4 vs 9.9 years); less likely to be female (62.9 vs 70.6%); more likely to have uveitis (12.7 vs 8.8%), asthma (9.1 vs 6.5%) or IBD (9.8 vs 1.4%); and more likely to have a hospitalization attributable to infection (3.1 vs 1.7%) in the 6 months preceding the index date. In contrast, the DMARD group had higher proportions of patients with a history of SLE (1.7 vs 0.4%), NSAID use (60.5 vs 42.7%), CS use (27.9 vs 23.4%) and antibiotic use (38.7 vs 31.1%). Table 1 Patient characteristics of DMARD and TNF inhibitor users Variables  DMARDs  TNFIs  P-value  (n = 2013)  (n = 482)  Patient characteristicsa        Age, mean (s.d.), years  9.9 (4.2)  10.4 (4.0)  0.018  Gender, n (%)            Male  592 (29.4)  179 (37.1)  0.001      Female  1421 (70.6)  303 (62.9)    Region, n (%)            Northeast  407 (20.2)  104 (21.6)  0.600      Midwest  530 (26.3)  111 (23.0)        South  661 (32.8)  170 (35.3)        West  366 (18.2)  86 (17.8)        Unknown  49 (2.4)  11 (2.3)    Capitated health plan type, n (%)            Non-capitalized plan  1575 (78.2)  379 (78.6)  0.582      Capitalized plan  296 (14.7)  75 (15.6)        Unknown  142 (7.1)  28 (5.8)    Year of medication use, n (%)            2009  277 (13.8)  52 (10.8)  0.007      2010  476 (23.6)  97 (20.1)        2011  509 (25.3)  110 (22.8)        2012  524 (26.0)  161 (33.4)        2013  227 (11.3)  62 (12.9)    Health-care utilizationb, n (%)        Number of outpatient visits            0–2  528 (26.2)  122 (25.3)  0.131      3–4  612 (30.4)  128 (26.6)        ≥5  873 (43.4)  232 (48.1)    Number of inpatient visits            0  1873 (93.0)  439 (91.1)  0.137      ≥1  140 (7.0)  43 (8.9)    Number of ED visits            0  1595 (79.2)  370 (76.8)  0.233      ≥1  418 (20.8)  112 (23.2)    RA surgery  16 (0.8)  1 (0.2)  0.223  Co-morbiditiesb, n (%)        Charlson Co-morbidity Index            0  1534 (76.2)  361 (74.9)  0.546      ≥1  479 (23.8)  121 (25.1)        Asthma  131 (6.5)  44 (9.1)  0.043      Diabetes  14 (0.7)  5 (1.0)  0.392      SLE  35 (1.7)  2 (0.4)  0.033      SS  8 (0.4)  0 (0.0)  0.367      Psoriasis  47 (2.3)  18 (3.7)  0.109      IBD  29 (1.4)  47 (9.8)  <0.0001      Uveitis  178 (8.8)  61 (12.7)  0.011  Medication useb, n (%)            NSAIDs  1217 (60.5)  206 (42.7)  <0.0001      CSs  561 (27.9)  113 (23.4)  0.049      Antibiotics  780 (38.7)  150 (31.1)  0.002      Anakinra  10 (0.5)  3 (0.6)  0.725      AZA  3 (0.1)  5 (1.0)  0.009      CYC  3 (0.1)  1 (0.2)  0.577      Mercaptopurine  4 (0.2)  6 (1.2)  0.005      Mycophenolate  2 (0.1)  0 (0.0)  1.000      Tacrolimus  5 (0.2)  2(0.4)  0.627  Previous infectionsb, n (%)            Any previous infections  849 (42.2)  229 (47.5)  0.034      Outpatient visit for infections  736 (36.6)  203 (42.1)  0.024      Hospitalizations for infections  34 (1.7)  15 (3.1)  0.043  Variables  DMARDs  TNFIs  P-value  (n = 2013)  (n = 482)  Patient characteristicsa        Age, mean (s.d.), years  9.9 (4.2)  10.4 (4.0)  0.018  Gender, n (%)            Male  592 (29.4)  179 (37.1)  0.001      Female  1421 (70.6)  303 (62.9)    Region, n (%)            Northeast  407 (20.2)  104 (21.6)  0.600      Midwest  530 (26.3)  111 (23.0)        South  661 (32.8)  170 (35.3)        West  366 (18.2)  86 (17.8)        Unknown  49 (2.4)  11 (2.3)    Capitated health plan type, n (%)            Non-capitalized plan  1575 (78.2)  379 (78.6)  0.582      Capitalized plan  296 (14.7)  75 (15.6)        Unknown  142 (7.1)  28 (5.8)    Year of medication use, n (%)            2009  277 (13.8)  52 (10.8)  0.007      2010  476 (23.6)  97 (20.1)        2011  509 (25.3)  110 (22.8)        2012  524 (26.0)  161 (33.4)        2013  227 (11.3)  62 (12.9)    Health-care utilizationb, n (%)        Number of outpatient visits            0–2  528 (26.2)  122 (25.3)  0.131      3–4  612 (30.4)  128 (26.6)        ≥5  873 (43.4)  232 (48.1)    Number of inpatient visits            0  1873 (93.0)  439 (91.1)  0.137      ≥1  140 (7.0)  43 (8.9)    Number of ED visits            0  1595 (79.2)  370 (76.8)  0.233      ≥1  418 (20.8)  112 (23.2)    RA surgery  16 (0.8)  1 (0.2)  0.223  Co-morbiditiesb, n (%)        Charlson Co-morbidity Index            0  1534 (76.2)  361 (74.9)  0.546      ≥1  479 (23.8)  121 (25.1)        Asthma  131 (6.5)  44 (9.1)  0.043      Diabetes  14 (0.7)  5 (1.0)  0.392      SLE  35 (1.7)  2 (0.4)  0.033      SS  8 (0.4)  0 (0.0)  0.367      Psoriasis  47 (2.3)  18 (3.7)  0.109      IBD  29 (1.4)  47 (9.8)  <0.0001      Uveitis  178 (8.8)  61 (12.7)  0.011  Medication useb, n (%)            NSAIDs  1217 (60.5)  206 (42.7)  <0.0001      CSs  561 (27.9)  113 (23.4)  0.049      Antibiotics  780 (38.7)  150 (31.1)  0.002      Anakinra  10 (0.5)  3 (0.6)  0.725      AZA  3 (0.1)  5 (1.0)  0.009      CYC  3 (0.1)  1 (0.2)  0.577      Mercaptopurine  4 (0.2)  6 (1.2)  0.005      Mycophenolate  2 (0.1)  0 (0.0)  1.000      Tacrolimus  5 (0.2)  2(0.4)  0.627  Previous infectionsb, n (%)            Any previous infections  849 (42.2)  229 (47.5)  0.034      Outpatient visit for infections  736 (36.6)  203 (42.1)  0.024      Hospitalizations for infections  34 (1.7)  15 (3.1)  0.043  a Patient characteristics were measured on the index date (i.e. new use of TNFIs or DMARDs). bThe covariates were measured in the 6 months before the index date. ED: emergency department; TNFI: TNF inhibitor. The mean follow-up time was 255 days for the DMARD group and 307 days for the TNFI group. We observed 18 and 11 serious infections in 1405.4 and 404.9 total person-years for the DMARD and TNFI groups, respectively; this resulted in crude rates of 1.28 (95% CI 0.76-2.02) and 2.72 (95%CI 1.36-4.86) serious infections per 100 person-years for these groups (Table 2). The TNFI group had 1.44 more serious infections per 100 person-years than the DMARD group. The crude rate ratio for infection was 2.12 (95% CI: 0.91, 4.74) for TNFIs compared with DMARDs. The higher rate of TNFI-associated serious infection was also reflected in the crude cumulative hazard estimates (log rank test P = 0.0357; Fig. 2). In addition, among the patients with infections, the median time to infection was 86 days (interquartile range 37–198) for the DMARD group and 91 days (interquartile range 21–207) for the TNFI group. Table 2 Risk of serious bacterial infections associated with TNF inhibitors compared with DMARDs Analysis  n  No. of serious infections  Crude rate of infection, events/100 person-years (95% CI)  Crude HR (95% CI)  Adjusted HR (95% CI)a  Main analysis                DMARDs  2013  18  1.28 (0.76, 2.02)  1.00 (reference)  1.00 (reference)      TNFIs  482  11  2.72 (1.36, 4.86)  2.27 (1.07, 4.80)  2.65 (1.06, 6.65)  Stratified analyses                Male                    DMARDs  592  3  0.74 (0.15, 2.17)  1.00 (reference)  1.00 (reference)          TNFIs  179  3  2.05 (0.42, 5.99)  2.91 (0.59, 14.50)  1.96 (0.32, 12.08)      Female                    DMARDs  1421  15  1.50 (0.84, 2.47)  1.00 (reference)  1.00 (reference)          TNFIs  303  8  3.10 (1.34, 6.10)  2.20 (0.93, 5.19)  2.92 (1.02, 8.32)  Analysis  n  No. of serious infections  Crude rate of infection, events/100 person-years (95% CI)  Crude HR (95% CI)  Adjusted HR (95% CI)a  Main analysis                DMARDs  2013  18  1.28 (0.76, 2.02)  1.00 (reference)  1.00 (reference)      TNFIs  482  11  2.72 (1.36, 4.86)  2.27 (1.07, 4.80)  2.65 (1.06, 6.65)  Stratified analyses                Male                    DMARDs  592  3  0.74 (0.15, 2.17)  1.00 (reference)  1.00 (reference)          TNFIs  179  3  2.05 (0.42, 5.99)  2.91 (0.59, 14.50)  1.96 (0.32, 12.08)      Female                    DMARDs  1421  15  1.50 (0.84, 2.47)  1.00 (reference)  1.00 (reference)          TNFIs  303  8  3.10 (1.34, 6.10)  2.20 (0.93, 5.19)  2.92 (1.02, 8.32)  a Adjusted for high-dimensional propensity score (tertile) in the model. HR: hazard ratio; TNFI: TNF inhibitor. Fig. 2 View largeDownload slide Three-year cumulative hazards of serious infections in children with JIA Fig. 2 View largeDownload slide Three-year cumulative hazards of serious infections in children with JIA Compared with DMARDs, use of TNFIs was associated with an increased risk of serious bacterial infection (HR = 2.74, 95% CI: 1.14, 6.56) after adjusting for the tertile of the hdPS (Table 2). In the gender-stratified analysis, an increased risk was observed (HR = 2.92, 95% CI: 1.02, 8.32) in female patients who initiated TNFIs. The point estimate was also elevated in the males (HR = 1.96, 95% CI: 0.32, 12.08) but was not statistically significant. Table 3 shows specific sites of infection in the TNFI and DMARD users. For both the TNFI and DMARD groups, infections were most commonly observed in the respiratory tract (36.4 and 33.3%), followed by the digestive system (27.3 and 22.2%), other organ systems (which included blood and device-related infections; 18.2 and 22.2%), skin and skin structure (9.1 and 16.7%) and genitourinary system (9.1 and 5.6%). Table 3 Site of infections among children with serious infection requiring hospitalizationa Site of infection  DMARDs  TNFIs  n (%)  n (%)  Number of patients with serious infection  18 (100.0)  11 (100.0)  Respiratory tract system  6 (33.3)  4 (36.4)      Upper respiratory tract infection  4 (22.2)  2 (18.2)      Pneumonia  2 (11.1)  2 (18.2)  Digestive system  4 (22.2)  3 (27.3)      Abdominal abscess  2 (11.1)  1 (9.1)      Cholecystitis  1 (5.6)  1 (9.1)      Gastroenteritis  1 (5.6)  1 (9.1)  Others  4 (22.2)  2 (18.2)      Bacteraemia/septicaemia  4 (22.2)  1 (9.1)      Device-associated infections  0 (0.0)  1 (9.1)  Skin and skin structure  3 (16.7)  1 (9.1)      Cellulitis  2 (11.1)  0 (0.0)      Necrotizing fasciitis  0 (0.0)  1 (9.1)      Septic arthritis  1 (5.6)  0 (0.0)  Genitourinary system  1 (5.6)  1 (9.1)      Pyelonephritis/urinary tract infection  1 (5.6)  1 (9.1)  Site of infection  DMARDs  TNFIs  n (%)  n (%)  Number of patients with serious infection  18 (100.0)  11 (100.0)  Respiratory tract system  6 (33.3)  4 (36.4)      Upper respiratory tract infection  4 (22.2)  2 (18.2)      Pneumonia  2 (11.1)  2 (18.2)  Digestive system  4 (22.2)  3 (27.3)      Abdominal abscess  2 (11.1)  1 (9.1)      Cholecystitis  1 (5.6)  1 (9.1)      Gastroenteritis  1 (5.6)  1 (9.1)  Others  4 (22.2)  2 (18.2)      Bacteraemia/septicaemia  4 (22.2)  1 (9.1)      Device-associated infections  0 (0.0)  1 (9.1)  Skin and skin structure  3 (16.7)  1 (9.1)      Cellulitis  2 (11.1)  0 (0.0)      Necrotizing fasciitis  0 (0.0)  1 (9.1)      Septic arthritis  1 (5.6)  0 (0.0)  Genitourinary system  1 (5.6)  1 (9.1)      Pyelonephritis/urinary tract infection  1 (5.6)  1 (9.1)  a Patients could have diagnoses of infections at more than one site at their event hospitalization, and the major site of infection was used based on clinical expert opinion. In the sensitivity analyses (Fig. 3), the risk estimate became non-significant when we restricted the study cohort to 2 to <16 years of age (HR = 2.53, 95% CI: 0.98, 6.55) and excluded patients with other rheumatoid diseases and IBD (HR = 1.15, 95% CI: 0.28, 4.17). However, we observed a greater infection risk for TNFIs when the gap for defining treatment discontinuation was reduced to 31 days (HR = 3.61, 95% CI: 1.32, 9.87) and when HCQ and SSZ users were excluded from the DMARD group (HR = 3.15, 95% CI: 1.14, 8.67). Fig. 3 View largeDownload slide Sensitivity analyses of the risk of serious infection associated with TNF inhibitors compared with DMARDs * indicates other rheumatic diseases including SLE, SS and psoriasis. ** indicates p value <0.05. DC: discontinuation; hdPS: high-dimensional propensity score. Filled circle represented adjusted HR of serious infection associated with TNFIs and the range represented 95% CI. Fig. 3 View largeDownload slide Sensitivity analyses of the risk of serious infection associated with TNF inhibitors compared with DMARDs * indicates other rheumatic diseases including SLE, SS and psoriasis. ** indicates p value <0.05. DC: discontinuation; hdPS: high-dimensional propensity score. Filled circle represented adjusted HR of serious infection associated with TNFIs and the range represented 95% CI. Discussion In a population of commercially insured children with JIA, we found that new use of TNFIs was associated with a 2.7-fold increase in risk of serious bacterial infection compared with new use of DMARDs. This increased risk estimate is consistent with the findings of some RA studies in adults [15, 16]. However, previous studies examining the association between TNFIs and infection in children with JIA had conflicting results. For example, Beukelman et al. [31, 32] analysed US national Medicaid data (2000–05) and found no difference between TNFIs and MTX in the rate of hospitalization for bacterial infections. Likewise, an analysis performed by Davies et al. [33] using data from the British Society for Paediatric and Adolescent Rheumatology Etanercept Cohort Study revealed no increased risk for serious infections requiring hospitalization and/or i.v. antibiotic use (adjusted HR 1.36, 95% CI: 0.60, 3.07) associated with etanercept compared with MTX. In contrast, Klotsche et al. [34] examined a set of German registries (2005–11) and found that the risk of infection leading to hospitalization was higher (relative risk 2.12, 95% CI: 1.08, 4.17) in patients receiving etanercept compared with MTX. Unlike previous studies, we adopted a new-user design for exposure groups. A recent study also examined new use of TNFIs and MTX in children with JIA from the Medicaid database [32]; however, in their study the use of MTX was considered a time-varying covariate, whereas in our study the effects of MTX were excluded from the TNFI group. Although TNFIs are usually recommended for use in combination with DMARDs, studies have shown that TNFIs are increasingly used early in the JIA disease course [35, 36] and as monotherapy [9, 36]. This new treatment paradigm facilitated our comparison of the two groups, as both consisted of new initiators of therapy. Notably, the black box warning for TNFIs stated that most infections developed when the drugs were used in combination with other immunosuppressants, such as DMARDs or CSs [37]. However, our study provides evidence that TNFI monotherapy is also associated with an increased risk of infection while controlling for CS use and other confounders. In addition, we further confirm the risk of infection associated with TNFI use in children with JIA. In our study, the rates of infection in the TNFI group (2.27/100 person-years) and DMARD group (1.28/100 person-years) were consistent with those observed in other studies of JIA [33, 38, 39]. The likely mechanism by which TNFIs increase the risk of infection is related to the role of TNF in the immune response [40]. Inhibiting the action of TNF prevents the normal immune response, increasing the risk for infections. It is possible that infections are not detected until they become severe because TNFIs impede the usual clinical presentation of high fever and elevated CRP. In addition, use of TNFIs may mimic a primary immunodeficiency that is considered a risk factor for mAb-related infections [41]. The time to infection (median of ∼90 days) was short in our study and was similar to findings of previous studies. For example, in the analysis by Davies et al. [33], 44% (24 of 54) of patients developed a serious infection requiring hospitalization within 6 months of etanercept initiation in children with JIA. A study of adults with RA also reported that the adjusted incidence rate ratio was 4.6 (95% CI: 1.8, 11.9) for TNFIs compared with DMARDs when restricting the follow-up time to the first 90 days [18]. These findings are consistent with the pharmacodynamic properties of TNFIs. The time to therapeutic effect of TNFIs varies by individual patient but is typically rapid; symptom improvement can be seen after two or three doses, and additional improvements over 3–6 months have been reported. Given these findings, health-care professionals should be vigilant in monitoring for symptoms of infection, especially in the first 3–6 months of treatment. Although our results suggest increased risk for infections associated with TNFIs compared with DMARDs, clinicians and patients should consider this risk in light of the benefits of TNFIs. Specifically, TNFIs are highly effective drugs that have been shown to improve symptoms, physical functioning, radiographic progression and quality of life [1–3]. In order to balance the risks and benefits associated with TNFIs, both the US FDA and the European Medicines Agency have developed risk-mitigation strategies for these biologics, but guidance on risk management for children is lacking [42, 43]. Our findings provide evidence that these agencies could use to adapt risk-management plans for children undergoing TNFI treatment. Such plans could incorporate appropriate screening, monitoring and even withholding of treatment to mitigate the potential harm of TNFIs to children with JIA. The strengths of our study include its use of relatively recent data (2009–13), use of a new-user design and use of an hdPS approach. However, our findings should be interpreted with consideration of the limitations. In particular, our results are subject to limitations common to studies using administrative claims databases, including potential misclassification of outcomes and exposure. However, we minimized this potential issue by using previously validated algorithms to identify infections and JIA [26, 27]. We also used comprehensive pharmacy records and intervention procedure codes to examine exposure, although we could not confirm that the patients administered the medications as directed. A dose–response relationship has been documented between use of CSs and higher risk of infection [44]. However, lack of information about children’s weight and prescribers’ tapering instructions resulted in considerable uncertainty in estimating the effect of the CS dose. In addition, because administrative data are not designed for research purposes, clinical data for health status in JIA (e.g. disease activity score and JIA subtype) were not available for analysis. Confounding from these and other unmeasured variables cannot be completely ruled out. Specifically, an important unmeasured confounder in our study is the relationship between JIA itself and infection. JIA or RA was found to be associated with an increased risk of infection [31, 45, 46], possibly because of T cell circulation and impaired thymic function [46, 47]. Our findings might have overestimated the TNFI–infection relationship if JIA severity was higher in the TNFI group. TNFIs are indicated for moderately to severely active polyarticular JIA, and thus children who receive TNFIs may have more severe JIA than those who receive DMARDs alone. Recognizing this possibility, we applied an hdPS approach using a large number of proxy indicators that indirectly reflected disease severity to account for the potential channelling bias. Several simulation studies have demonstrated the effectiveness of this method [28, 29]. Nonetheless, residual confounding of the JIA–infection relationship might remain. Another potential limitation of our study is insufficient statistical power. Children with JIA constitute a smaller population than adults with RA. Although the rate of TNFI-related infection in our study was not extremely low, it is still challenging for any study to obtain a large enough sample size of children with JIA to examine safety outcomes adequately, especially in subgroup analyses. As a result, we could not perform analyses to evaluate the infection risk for individual TNFI agents and for combination therapy using TNFIs and CSs, nor could we conduct analyses using only the primary diagnosis in the inpatient claims for identifying patients with severe infection. In our analysis excluding patients with other rheumatic diseases and IBD, the risk of infection reduced toward the null. On the one hand, it is possible that IBD might confound the relationship between TNFIs and serious infection, because IBD patients were more likely to receive TNFI monotherapy and develop infections. On the other hand, the small number of observed infections yielded a wide 95% CI (0.28, 4.17), and we could not draw a conclusion with insufficient power in this analysis. Future studies are needed to examine the confounding effect of IBD and confirm our findings. To facilitate studies of drug safety for children, it would be beneficial to incorporate relevant data into a national- or international-scale surveillance system. As one example, the European registry called Pharmachild is a pharmacovigilance project that aims to observe long-term adverse events associated with use of DMARDs and biologics in children with JIA across 50 participating countries [48]. In the USA, the Childhood Arthritis and Rheumatology Research Alliance has initiated a prospective observational registry enrolling patients with paediatric rheumatic diseases across North America for research purposes, including research on long-term safety of medications [49]. Also, the FDA Sentinel Initiative is a national electronic surveillance system designed to monitor proactively and examine the safety of medications and biologics [50]. In future studies, use of data from such registries and systems would support a more thorough examination of potentially serious adverse events related to TNFIs. Finally, the FDA-required black box warning might have impacted the observed association between TNFIs and infection. For example, physicians might have checked for infections more often in TNFI users than in DMARD users. In addition, the warning might have impacted the decision to admit to the hospital a patient with suspected infection differently depending on whether they were on TNFIs or DMARDs. Therefore, a differential detection or management bias might exist between the two exposure groups, which might have resulted in overestimation of relative risk in the TNFI group. In contrast, the FDA warning might have caused prescribers to avoid use of TNFIs for children susceptible to infection, and these children might have been channelled to DMARD treatment. In this case, our findings might have underestimated the relative risk in the TNFI group. However, we attempted to control for this type of confounding by applying contemporary analytical methods that included the hdPS modelling. In summary, our study demonstrated a higher risk of serious bacterial infection associated with use of TNFIs compared with DMARDs in children with JIA. Our analysis supports the FDA warning about TNFI-associated infection in children with JIA and also provides a comparison between DMARD and TNFI monotherapy. Future studies incorporating a larger cohort of children with JIA would help to confirm our findings and further characterize the risk of infection across individual TNFI agents. In the meantime, clinicians and patients need to balance the benefits of these highly effective drugs against the risk for infection they pose. Acknowledgements We gratefully acknowledge the editorial support of Mr Jon Mann, Academic Counsellor, University of Illinois at Chicago, during the preparation of this manuscript. The views expressed in this article are those of the authors and do not necessarily represent the views of the Department of Veterans Affairs or Health Services Research and Development Service. Funding: No specific funding was received from any bodies in the public, commercial or not-for-profit sectors to carry out the work described in this manuscript. Disclosure statement: G.T.S. has received payment for consulting or speaking from AbbVie Inc., Astellas Pharma US Inc. and Baxter International Inc., in the past 3 years. All other authors have declared no conflicts of interest. Supplementary data Supplementary data are available at Rheumatology Online. References 1 Lovell DJ, Giannini EH, Reiff A et al.  ; Pediatric Rheumatology Collaborative Study Group. Etanercept in children with polyarticular juvenile rheumatoid arthritis. N Engl J Med  2000; 342: 763– 9. Google Scholar CrossRef Search ADS PubMed  2 Lovell DJ, Ruperto N, Goodman S et al.   Adalimumab with or without methotrexate in juvenile rheumatoid arthritis. N Engl J Med  2008; 359: 810– 20. Google Scholar CrossRef Search ADS PubMed  3 Keystone EC, Kavanaugh AF, Sharp JT et al.   Radiographic, clinical, and functional outcomes of treatment with adalimumab (a human anti-tumor necrosis factor monoclonal antibody) in patients with active rheumatoid arthritis receiving concomitant methotrexate therapy: a randomized, placebo-controlled, 52-week trial. Arthritis Rheum  2004; 50: 1400– 11. Google Scholar CrossRef Search ADS PubMed  4 Beukelman T, Patkar NM, Saag KG et al.   2011 American College of Rheumatology recommendations for the treatment of juvenile idiopathic arthritis: initiation and safety monitoring of therapeutic agents for the treatment of arthritis and systemic features. Arthritis Care Res  2011; 63: 465– 82. Google Scholar CrossRef Search ADS   5 Singh JA, Saag KG, Bridges SLJr et al.   2015 American College of Rheumatology Guideline for the treatment of rheumatoid arthritis. Arthritis Rheumatol  2016; 68: 1– 26. Google Scholar PubMed  6 Breedveld FC, Emery P, Keystone E et al.   Infliximab in active early rheumatoid arthritis. Ann Rheum Dis  2004; 63: 149– 55. Google Scholar CrossRef Search ADS PubMed  7 St Clair EW, van der Heijde DM, Smolen JS et al.   Combination of infliximab and methotrexate therapy for early rheumatoid arthritis: a randomized, controlled trial. Arthritis Rheum  2004; 50: 3432– 43. Google Scholar CrossRef Search ADS PubMed  8 Lee SJ, Chang H, Yazici Y et al.   Utilization trends of tumor necrosis factor inhibitors among patients with rheumatoid arthritis in a United States observational cohort study. J Rheumatol  2009; 36: 1611– 7. Google Scholar CrossRef Search ADS PubMed  9 Mannion ML, Xie F, Curtis JR, Beukelman T. Recent trends in medication usage for the treatment of juvenile idiopathic arthritis and the influence of tumor necrosis factor inhibitors. J Rheumatol  2014; 41: 2078– 84. Google Scholar CrossRef Search ADS PubMed  10 Schneeweiss S, Setoguchi S, Weinblatt ME et al.   Anti-tumor necrosis factor alpha therapy and the risk of serious bacterial infections in elderly patients with rheumatoid arthritis. Arthritis Rheum  2007; 56: 1754– 64. Google Scholar CrossRef Search ADS PubMed  11 Grijalva CG, Chen L, Delzell E et al.   Initiation of tumor necrosis factor-α antagonists and the risk of hospitalization for infection in patients with autoimmune diseases. JAMA  2011; 306: 2331– 9. Google Scholar CrossRef Search ADS PubMed  12 Bongartz T, Sutton AJ, Sweeting MJ et al.   Anti-TNF antibody therapy in rheumatoid arthritis and the risk of serious infections and malignancies: systematic review and meta-analysis of rare harmful effects in randomized controlled trials. JAMA  2006; 295: 2275– 85. Google Scholar CrossRef Search ADS PubMed  13 Bernatsky S, Habel Y, Rahme E. Observational studies of infections in rheumatoid arthritis: a metaanalysis of tumor necrosis factor antagonists. J Rheumatol  2010; 37: 928– 31. Google Scholar CrossRef Search ADS PubMed  14 Alonso-Ruiz A, Pijoan JI, Ansuategui E et al.   Tumor necrosis factor alpha drugs in rheumatoid arthritis: systematic review and metaanalysis of efficacy and safety. BMC Musculoskelet Disord  2008; 9: 52. Google Scholar CrossRef Search ADS PubMed  15 Curtis JR, Patkar N, Xie A et al.   Risk of serious bacterial infections among rheumatoid arthritis patients exposed to tumor necrosis factor α antagonists. Arthritis Rheum  2007; 56: 1125– 33. Google Scholar CrossRef Search ADS PubMed  16 Lane MA, McDonald JR, Zeringue AL et al.   TNF-α antagonist use and risk of hospitalization for infection in a national cohort of veterans with rheumatoid arthritis. Medicine  2011; 90: 139– 45. Google Scholar CrossRef Search ADS PubMed  17 Listing J, Strangfeld A, Kary S et al.   Infections in patients with rheumatoid arthritis treated with biologic agents. Arthritis Rheum  2005; 52: 3403– 12. Google Scholar CrossRef Search ADS PubMed  18 Dixon WG, Symmons DPM, Lunt M et al.   Serious infection following anti-tumor necrosis factor α therapy in patients with rheumatoid arthritis: lessons from interpreting data from observational studies. Arthritis Rheum  2007; 56: 2896– 904. Google Scholar CrossRef Search ADS PubMed  19 Salmon-Ceron D, Tubach F, Lortholary O et al.   Drug-specific risk of non-tuberculosis opportunistic infections in patients receiving anti-TNF therapy reported to the 3-year prospective French RATIO registry. Ann Rheum Dis  2011; 70: 616– 23. Google Scholar CrossRef Search ADS PubMed  20 Information for Healthcare Professionals: Cimzia (certolizumab pegol), Enbrel (etanercept), Humira (adalimumab), and Remicade (infliximab). http://www.fda.gov/Drugs/DrugSafety/PostmarketDrugSafetyInformationforPatientsandProviders/ucm124185.htm. (Accessed on 12 November 12, 2016, date last accessed). 21 Kaushal R, Bates DW, Landrigan C et al.   Medication errors and adverse drug events in pediatric inpatients. JAMA  2001; 285: 2114– 20. Google Scholar CrossRef Search ADS PubMed  22 World Health Organization. Promoting Safety of Medicinces For Children. 2007. http://www.who.int/medicines/publications/essentialmedicines/Promotion_safe_med_childrens.pdf (12 November 2016, date last accessed). 23 Ruperto N, Lovell DJ, Cuttica R et al.   A randomized, placebo-controlled trial of infliximab plus methotrexate for the treatment of polyarticular-course juvenile rheumatoid arthritis. Arthritis Rheum  2007; 56: 3096– 106. Google Scholar CrossRef Search ADS PubMed  24 Horneff G, De Bock F, Foeldvari I et al.   Safety and efficacy of combination of etanercept and methotrexate compared to treatment with etanercept only in patients with juvenile idiopathic arthritis (JIA): preliminary data from the German JIA Registry. Ann Rheum Dis  2009; 68: 519– 25. Google Scholar CrossRef Search ADS PubMed  25 Danielson E. White Paper. Health Research Data for the Real World: The MarketScan® Databases. 2011. http://truvenhealth.com/portals/0/assets/PH_11238_0612_TEMP_MarketScan_WP_FINAL.pdf (23 March 2016, date last accessed). 26 Harrold LR, Salman C, Shoor S et al.   Incidence and prevalence of juvenile idiopathic arthritis among children in a managed care population, 1996–2009. J Rheumatol  2013; 40: 1218– 25. Google Scholar CrossRef Search ADS PubMed  27 Patkar NM, Curtis JR, Teng GG et al.   Administrative codes combined with medical records based criteria accurately identified bacterial infections among rheumatoid arthritis patients. J Clin Epidemiol  2009; 62: 321– 7, 327 e1–7. Google Scholar CrossRef Search ADS PubMed  28 Schneeweiss S, Rassen JA, Glynn RJ et al.   High-dimensional propensity score adjustment in studies of treatment effects using health care claims data. Epidemiology  2009; 20: 512– 22. Google Scholar CrossRef Search ADS PubMed  29 Rassen JA, Glynn RJ, Brookhart MA, Schneeweiss S. Covariate selection in high-dimensional propensity score analyses of treatment effects in small samples. Am J Epidemiol  2011; 173: 1404– 13. Google Scholar CrossRef Search ADS PubMed  30 Pharmacodynamics and Pharmacokinetics section. Drug information. UpToDate. http://www.uptodate.com (20 July 2016, date last accessed). 31 Beukelman T, Xie F, Chen L et al.   Rates of hospitalized bacterial infection associated with juvenile idiopathic arthritis and its treatment. Arthritis Rheum  2012; 64: 2773– 80. Google Scholar CrossRef Search ADS PubMed  32 Beukelman T, Xie F, Baddley JW et al.   The risk of hospitalized infection following initiation of biologic agents versus methotrexate in the treatment of juvenile idiopathic arthritis. Arthritis Res Ther  2016; 18: 210. Google Scholar CrossRef Search ADS PubMed  33 Davies R, Southwood TR, Kearsley-Fleet L et al.   Medically significant infections are increased in patients with juvenile idiopathic arthritis treated with etanercept: results from the British Society for Paediatric and Adolescent Rheumatology Etanercept Cohort Study. Arthritis Rheumatol  2015; 67: 2487– 94. Google Scholar CrossRef Search ADS PubMed  34 Klotsche J, Niewerth M, Haas JP et al.   Long-term safety of etanercept and adalimumab compared to methotrexate in patients with juvenile idiopathic arthritis (JIA). Ann Rheum Dis  2016; 75: 855– 61. Google Scholar CrossRef Search ADS PubMed  35 Otten MH, Anink J, Prince FH et al.   Trends in prescription of biological agents and outcomes of juvenile idiopathic arthritis: results of the Dutch national Arthritis and Biologics in Children Register. Ann Rheum Dis  2015; 74: 1379– 86. Google Scholar CrossRef Search ADS PubMed  36 Lee WJ, Briars L, Lee TA et al.   Use of tumor necrosis factor-alpha inhibitors in children and young adults with juvenile idiopathic arthritis or rheumatoid arthritis. Pharmacotherapy  2016; 36: 1201– 9. Google Scholar CrossRef Search ADS PubMed  37 Product Information: ENBREL (etanercept) for subcutaneous injection. Amgen, Inc., Thousand Oaks, CA. 2016. http://www.accessdata.fda.gov/drugsatfda_docs/label/2015/ 103795s5548lbl.pdf (8 March 2017, date last accessed). 38 Lovell DJ, Reiff A, Ilowite NT et al.   Safety and efficacy of up to eight years of continuous etanercept therapy in patients with juvenile rheumatoid arthritis. Arthritis Rheum  2008; 58: 1496– 504. Google Scholar CrossRef Search ADS PubMed  39 Tarkiainen M, Tynjälä P, Vähäsalo P, Lahdenne P. Occurrence of adverse events in patients with JIA receiving biologic agents: long-term follow-up in a real-life setting. Rheumatology  2015; 54: 1170– 6. Google Scholar CrossRef Search ADS PubMed  40 Wajant H, Pfizenmaier K, Scheurich P. Tumor necrosis factor signaling. Cell Death Differ  2003; 10: 45– 65. Google Scholar CrossRef Search ADS PubMed  41 Maródi L, Casanova JL. Primary immunodeficiencies may reveal potential infectious diseases associated with immune-targeting mAb treatments. J Allergy Clin Immunol  2010; 126: 910– 7. Google Scholar CrossRef Search ADS PubMed  42 European Medicines Agency. http://www.ema.europa.eu/ema (5 July 2016, date last accessed). 43 Approved Risk Evaluation and Mitigation Strategies (REMS). US Food and Drug Administration. http://www.accessdata.fda.gov/scripts/cder/rems (5 July 2016, date last accessed). 44 Lovell DJ, Zaoutis TE, Sullivan K. Immunosuppressants, infection, and inflammation. Clin Immunol  2004; 113: 137– 9. Google Scholar CrossRef Search ADS PubMed  45 Smitten AL, Choi HK, Hochberg MC et al.   The risk of hospitalized infection in patients with rheumatoid arthritis. J Rheumatol  2008; 35: 387– 93. Google Scholar PubMed  46 Blumentals WA, Arreglado A, Napalkov P, Toovey S. Rheumatoid arthritis and the incidence of influenza and influenza-related complications: a retrospective cohort study. BMC Musculoskelet Disord  2012; 13: 158. Google Scholar CrossRef Search ADS PubMed  47 Koetz K, Bryl E, Spickschen K et al.   T cell homeostasis in patients with rheumatoid arthritis. Proc Natl Acad Sci USA  2000; 97: 9203– 8. Google Scholar CrossRef Search ADS PubMed  48 Swart J, Consolaro A, Horneff G et al.   Pharmacovigilance in juvenile idiopathic arthritis patients (Pharmachild) treated with biologic agents and/or methotrexate. Consolidated baseline characteristics from Pharmachild and other national registries. Pediatr Rheumatol Online J  2014; 12: P7. Google Scholar CrossRef Search ADS   49 The Childhood Arthritis and Rheumatology Research Alliance (CARRA) registry. https://carragroup.org/carra-registry-updates/ (2 November 2016, date last accessed). 50 FDA's Sentinel Initiative. http://www.fda.gov/Safety/FDAsSentinelInitiative/ucm2007250.htm (20 June 2016, date last accessed). © The Author 2017. Published by Oxford University Press on behalf of the British Society for Rheumatology. All rights reserved. For Permissions, please email: journals.permissions@oup.com http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Rheumatology Oxford University Press

Risk of serious bacterial infection associated with tumour necrosis factor-alpha inhibitors in children with juvenile idiopathic arthritis

Loading next page...
 
/lp/ou_press/risk-of-serious-bacterial-infection-associated-with-tumour-necrosis-c65edfMZCd
Publisher
Oxford University Press
Copyright
© The Author 2017. Published by Oxford University Press on behalf of the British Society for Rheumatology. All rights reserved. For Permissions, please email: journals.permissions@oup.com
ISSN
1462-0324
eISSN
1462-0332
D.O.I.
10.1093/rheumatology/kex049
Publisher site
See Article on Publisher Site

Abstract

Abstract Objectives TNF-α inhibitors (TNFIs) have a black box warning for increased risk of serious infection that was based on evidence from studies of adults. Evidence of the association is lacking for children. We aimed to examine the risk of infection posed by TNFIs compared with DMARDs in children with JIA. Methods We conducted a cohort study using the 2009–13 Truven MarketScan Commercial Claims and Encounters database. Children <16 years old with JIA who initiated monotherapy with TNFIs or DMARDs were identified and followed for occurrence of serious bacterial infection requiring hospitalization. Cox proportional hazard models were used to estimate hazard ratios for infection associated with TNFIs compared with DMARDs, adjusting for potential confounders with high-dimensional propensity scores and time-varying CS use. Results We identified 2013 DMARD initiators and 482 TNFI initiators with a mean follow-up of 255 and 307 days, respectively. We identified 18 and 11 patients with a serious infection in the DMARD and TNFI groups, resulting in crude rates of 1.28 (95% CI 0.76-2.02) and 2.72 (95%CI 1.36-4.86) per 100 person-years, respectively. In adjusted models, TNFIs were associated with an increased risk of serious bacterial infection compared with DMARDs (adjusted hazard ratio 2.72, 95% CI: 1.08, 6.86). Conclusion Use of TNFIs poses a higher risk of serious infection compared with DMARDs in children with JIA. Our analysis confirms the US Food and Drug Administration warning about TNFI-associated infection in children with JIA. biologic therapies, tumour necrosis factor-alpha inhibitors, juvenile idiopathic arthritis, bacterial infection, children, drug safety Rheumatology key messages TNF inhibitors pose higher risk of serious bacterial infection than DMARDs in children with JIA. The most common site for TNF inhibitor-associated infection in JIA is the respiratory tract. The US Food and Drug Administration warning about TNF inhibitor-associated infection is confirmed in children with JIA Introduction TNF-α inhibitors (TNFIs), including mAb (e.g. infliximab and adalimumab) and fusion proteins (etanercept), are biological agents used to treat RA and JIA. TNFIs are highly effective for RA and JIA and have been shown in clinical trials to induce disease remission, improve quality of life and enhance physical functioning [1–3]. The ACR recommends use of TNFIs for patients who have received MTX or LEF for 3–6 months and whose disease activity remains uncontrolled [4, 5]. In addition, earlier use of TNFIs has been shown to improve short-term clinical outcomes in both RA and JIA [6, 7]. As a result, TNFI use has increased over the past decade [8, 9]. Although TNFIs are generally safe, growth in their use has resulted in more reports of adverse events. In particular, infections have been the most frequently reported serious adverse event in adult RA patients [10, 11]. However, mixed findings were observed in evaluations of the association between TNFIs and infection in meta-analyses [12–14] and observational studies [10, 11, 15–19]. Some studies reported increased risk of infection for TNFIs compared with DMARDs [12, 13, 15, 16, 19], whereas others found no elevated risk [10, 11, 14, 17, 18]. Nevertheless, in 2008, the US Food and Drug Administration (FDA) required TNFI manufacturers to include a black box warning in the product label for serious infections leading to hospitalization or death [20]. The warning applies to both adults and children, yet most studies evaluating the TNFI–infection association were conducted in adults with RA. Children are a vulnerable population more likely to experience adverse drug events than adults [21]. In addition, the immaturity of their immune system may put children at a higher risk for infection [22]. Unfortunately, clinical and observational studies involving JIA patients have had limited numbers of participants and short follow-up periods [1, 2, 23, 24]. More definitive evidence is needed for the association between TNFIs and infection in children. The aim of this study was to examine the risk of serious bacterial infection associated with TNFIs in children with JIA. Methods Data sources In this retrospective cohort study, we analysed data from the Truven Health MarketScan Commercial Claims and Encounters database for the period from 1 January 2009 to 31 December 2013. The database contains employer-based health insurance claims for enrollees and their dependants across the USA. Administrative data on patient enrolment; health-care utilization, including hospitalizations and visits to outpatient clinics and emergency departments; medical procedures; costs of services; and pharmacy records are available in the database and for research use [25]. Ethical approval was not required to analyse the data in this study. Population The analytical cohort was developed by first identifying children (age <16 years) with JIA who had at least one prescription for a TNFI or DMARD during the study period. JIA was identified using a previously validated algorithm that included International Classification of Disease, ninth version, Clinical Modification codes 714.xx (RA and JIA), 696.0 (PsA) and 720.xx (AS) [26]. Included children were required to have two or more JIA diagnoses within 1 year or one JIA diagnosis coded by a paediatrician or rheumatologist [26]. The date of the first new prescription for a TNFI or DMARD was defined as the index date and included new users who had to have no previous use of either TNFIs or DMARDs 6 months before the index date. In addition, patients were excluded if they met any of the following criteria during the 6 months before the index date: (i) discontinued health plan enrolment; (ii) history of tuberculosis and/or use of medications for tuberculosis; or (iii) history of cancer, transplantation and/or HIV infection. Exposures Exposures to TNFIs and DMARDs were identified using National Drug Codes and Healthcare Common Procedure Coding System. DMARDs included MTX, HCQ, SSZ and LEF, and TNFIs included etanercept, adalimumab, infliximab, certolizumab and golimumab. Each included patient was followed from the index date to the first occurrence of infection, disenrolment from the health plan, discontinuation of treatment, a switch between or a combined use of a TNFI and DMARD, or the end of the study period (31 December 2013). The observation time was the time of continuous use of TNFI monotherapy or DMARD monotherapy. Discontinuation of medication was defined as a gap >92 days between the end of one prescription’s supply (in days) and the next prescription date. For patients who discontinued their medication, we censored them at the end of the last prescription’s days supply before a gap >92 days. Outcomes The outcome of interest was the first occurrence of a serious bacterial infection, which was defined as an infection requiring hospitalization. We applied a previously validated algorithm that incorporated 27 infection sites using International Classification of Disease, ninth version, Clinical Modification codes at any diagnosis position in the inpatient claims [27]. If an individual was diagnosed as having an infection at two or more sites on the same date, the major site of infection was identified based on the severity of infection, concomitant diagnoses and procedures recorded. Only 14% of cases of serious infections had more than one site of infection. High-dimensional propensity score models We used a high-dimensional propensity score (hdPS) to identify and adjust for a large number of covariates that could confound the association between the exposure and infections [28, 29]. Variables were grouped in dimensions that included outpatient diagnoses, inpatient diagnoses, outpatient procedures, inpatient procedures and outpatient medication use during the 6 months before the index date. Thus, any medical conditions (including previous infection) represented by diagnosis or procedure codes or medication use were examined using the hdPS algorithm. The algorithm calculated and ranked a measure of confounding bias for each variable and selected the 500 top-ranked variables into the propensity score model (see supplementary Table S1, available at Rheumatology Online). Along with the 500 empirical variables, we included demographic variables (age, gender, geographical location, calendar year of medication use and type of health plan on the index date) and health-care utilization variables (the number of visits to outpatient, inpatient and emergency departments during the 6 months before the index date) in the logistic regression model to calculate the probability of receiving TNFIs (i.e. the propensity score). Statistical analysis Demographic information, health-care utilization, co-morbidities and concomitant medication use were compared between the TNFI and DMARD groups using Student’s unpaired t-test, a χ2 test or Fisher’s exact test, as appropriate. Crude rates of serious infection (number of events per 100 person-years) and 95% CIs were computed using the Poisson exact method. Specific infection sites were described. A Kaplan–Meier plot and log-rank test were used to examine the rate of serious infection over time. Cox proportional hazard models were used to estimate the hazard ratios (HRs) for serious infection associated with use of TNFIs compared with DMARDs. The tertile of the propensity score was included in the final Cox model. Analyses stratified by gender were also performed. We also conducted several sensitivity analyses to examine the robustness of our findings, as follows: (i) we adjusted for CS use in a time-varying manner to account for the effect of CSs during follow-up; (ii) we restricted our analysis to children ⩾2 to <16 years of age, which is consistent with the FDA label; (iii) we extended the observation period by 30 and 90 days beyond discontinuation of medication use to capture infections occurring after discontinuation; (iv) we reduced the gap defining treatment discontinuation from 92 to 31 days based on the duration of drug effect [1, 30]; (v) We excluded patients with SLE, SS or psoriasis 6 months before the index date owing to the fact that patients with autoimmune diseases may have different baseline risk for infection; (vi) we also excluded patients with IBD from (v); (vii) we excluded patients who used HCQ or SSZ in the DMARD group because users of these two medications usually had less severe disease; and (viii) we excluded 1% and 5% of patients with extreme hdPS, respectively. All analyses were performed using SAS statistical software version 9.4 (SAS Institute, Cary, NC, USA) and STATA 12 (StataCorp, College Station, TX, USA). The Institutional Review Board determined this study to be non-human subject research. Results We identified 5497 children with JIA who were prescribed either a TNFI or a DMARD during the study period (Fig. 1). After applying the exclusion criteria, the final study cohort consisted of 2495 individuals, including 2013 new DMARD users and 482 new TNFI users. Fig. 1 View largeDownload slide Selection criteria for analytical study cohort Fig. 1 View largeDownload slide Selection criteria for analytical study cohort Baseline characteristics were compared between the TNFI and DMARD groups (Table 1). TNFI initiators were slightly older than DMARD initiators (mean age: 10.4 vs 9.9 years); less likely to be female (62.9 vs 70.6%); more likely to have uveitis (12.7 vs 8.8%), asthma (9.1 vs 6.5%) or IBD (9.8 vs 1.4%); and more likely to have a hospitalization attributable to infection (3.1 vs 1.7%) in the 6 months preceding the index date. In contrast, the DMARD group had higher proportions of patients with a history of SLE (1.7 vs 0.4%), NSAID use (60.5 vs 42.7%), CS use (27.9 vs 23.4%) and antibiotic use (38.7 vs 31.1%). Table 1 Patient characteristics of DMARD and TNF inhibitor users Variables  DMARDs  TNFIs  P-value  (n = 2013)  (n = 482)  Patient characteristicsa        Age, mean (s.d.), years  9.9 (4.2)  10.4 (4.0)  0.018  Gender, n (%)            Male  592 (29.4)  179 (37.1)  0.001      Female  1421 (70.6)  303 (62.9)    Region, n (%)            Northeast  407 (20.2)  104 (21.6)  0.600      Midwest  530 (26.3)  111 (23.0)        South  661 (32.8)  170 (35.3)        West  366 (18.2)  86 (17.8)        Unknown  49 (2.4)  11 (2.3)    Capitated health plan type, n (%)            Non-capitalized plan  1575 (78.2)  379 (78.6)  0.582      Capitalized plan  296 (14.7)  75 (15.6)        Unknown  142 (7.1)  28 (5.8)    Year of medication use, n (%)            2009  277 (13.8)  52 (10.8)  0.007      2010  476 (23.6)  97 (20.1)        2011  509 (25.3)  110 (22.8)        2012  524 (26.0)  161 (33.4)        2013  227 (11.3)  62 (12.9)    Health-care utilizationb, n (%)        Number of outpatient visits            0–2  528 (26.2)  122 (25.3)  0.131      3–4  612 (30.4)  128 (26.6)        ≥5  873 (43.4)  232 (48.1)    Number of inpatient visits            0  1873 (93.0)  439 (91.1)  0.137      ≥1  140 (7.0)  43 (8.9)    Number of ED visits            0  1595 (79.2)  370 (76.8)  0.233      ≥1  418 (20.8)  112 (23.2)    RA surgery  16 (0.8)  1 (0.2)  0.223  Co-morbiditiesb, n (%)        Charlson Co-morbidity Index            0  1534 (76.2)  361 (74.9)  0.546      ≥1  479 (23.8)  121 (25.1)        Asthma  131 (6.5)  44 (9.1)  0.043      Diabetes  14 (0.7)  5 (1.0)  0.392      SLE  35 (1.7)  2 (0.4)  0.033      SS  8 (0.4)  0 (0.0)  0.367      Psoriasis  47 (2.3)  18 (3.7)  0.109      IBD  29 (1.4)  47 (9.8)  <0.0001      Uveitis  178 (8.8)  61 (12.7)  0.011  Medication useb, n (%)            NSAIDs  1217 (60.5)  206 (42.7)  <0.0001      CSs  561 (27.9)  113 (23.4)  0.049      Antibiotics  780 (38.7)  150 (31.1)  0.002      Anakinra  10 (0.5)  3 (0.6)  0.725      AZA  3 (0.1)  5 (1.0)  0.009      CYC  3 (0.1)  1 (0.2)  0.577      Mercaptopurine  4 (0.2)  6 (1.2)  0.005      Mycophenolate  2 (0.1)  0 (0.0)  1.000      Tacrolimus  5 (0.2)  2(0.4)  0.627  Previous infectionsb, n (%)            Any previous infections  849 (42.2)  229 (47.5)  0.034      Outpatient visit for infections  736 (36.6)  203 (42.1)  0.024      Hospitalizations for infections  34 (1.7)  15 (3.1)  0.043  Variables  DMARDs  TNFIs  P-value  (n = 2013)  (n = 482)  Patient characteristicsa        Age, mean (s.d.), years  9.9 (4.2)  10.4 (4.0)  0.018  Gender, n (%)            Male  592 (29.4)  179 (37.1)  0.001      Female  1421 (70.6)  303 (62.9)    Region, n (%)            Northeast  407 (20.2)  104 (21.6)  0.600      Midwest  530 (26.3)  111 (23.0)        South  661 (32.8)  170 (35.3)        West  366 (18.2)  86 (17.8)        Unknown  49 (2.4)  11 (2.3)    Capitated health plan type, n (%)            Non-capitalized plan  1575 (78.2)  379 (78.6)  0.582      Capitalized plan  296 (14.7)  75 (15.6)        Unknown  142 (7.1)  28 (5.8)    Year of medication use, n (%)            2009  277 (13.8)  52 (10.8)  0.007      2010  476 (23.6)  97 (20.1)        2011  509 (25.3)  110 (22.8)        2012  524 (26.0)  161 (33.4)        2013  227 (11.3)  62 (12.9)    Health-care utilizationb, n (%)        Number of outpatient visits            0–2  528 (26.2)  122 (25.3)  0.131      3–4  612 (30.4)  128 (26.6)        ≥5  873 (43.4)  232 (48.1)    Number of inpatient visits            0  1873 (93.0)  439 (91.1)  0.137      ≥1  140 (7.0)  43 (8.9)    Number of ED visits            0  1595 (79.2)  370 (76.8)  0.233      ≥1  418 (20.8)  112 (23.2)    RA surgery  16 (0.8)  1 (0.2)  0.223  Co-morbiditiesb, n (%)        Charlson Co-morbidity Index            0  1534 (76.2)  361 (74.9)  0.546      ≥1  479 (23.8)  121 (25.1)        Asthma  131 (6.5)  44 (9.1)  0.043      Diabetes  14 (0.7)  5 (1.0)  0.392      SLE  35 (1.7)  2 (0.4)  0.033      SS  8 (0.4)  0 (0.0)  0.367      Psoriasis  47 (2.3)  18 (3.7)  0.109      IBD  29 (1.4)  47 (9.8)  <0.0001      Uveitis  178 (8.8)  61 (12.7)  0.011  Medication useb, n (%)            NSAIDs  1217 (60.5)  206 (42.7)  <0.0001      CSs  561 (27.9)  113 (23.4)  0.049      Antibiotics  780 (38.7)  150 (31.1)  0.002      Anakinra  10 (0.5)  3 (0.6)  0.725      AZA  3 (0.1)  5 (1.0)  0.009      CYC  3 (0.1)  1 (0.2)  0.577      Mercaptopurine  4 (0.2)  6 (1.2)  0.005      Mycophenolate  2 (0.1)  0 (0.0)  1.000      Tacrolimus  5 (0.2)  2(0.4)  0.627  Previous infectionsb, n (%)            Any previous infections  849 (42.2)  229 (47.5)  0.034      Outpatient visit for infections  736 (36.6)  203 (42.1)  0.024      Hospitalizations for infections  34 (1.7)  15 (3.1)  0.043  a Patient characteristics were measured on the index date (i.e. new use of TNFIs or DMARDs). bThe covariates were measured in the 6 months before the index date. ED: emergency department; TNFI: TNF inhibitor. The mean follow-up time was 255 days for the DMARD group and 307 days for the TNFI group. We observed 18 and 11 serious infections in 1405.4 and 404.9 total person-years for the DMARD and TNFI groups, respectively; this resulted in crude rates of 1.28 (95% CI 0.76-2.02) and 2.72 (95%CI 1.36-4.86) serious infections per 100 person-years for these groups (Table 2). The TNFI group had 1.44 more serious infections per 100 person-years than the DMARD group. The crude rate ratio for infection was 2.12 (95% CI: 0.91, 4.74) for TNFIs compared with DMARDs. The higher rate of TNFI-associated serious infection was also reflected in the crude cumulative hazard estimates (log rank test P = 0.0357; Fig. 2). In addition, among the patients with infections, the median time to infection was 86 days (interquartile range 37–198) for the DMARD group and 91 days (interquartile range 21–207) for the TNFI group. Table 2 Risk of serious bacterial infections associated with TNF inhibitors compared with DMARDs Analysis  n  No. of serious infections  Crude rate of infection, events/100 person-years (95% CI)  Crude HR (95% CI)  Adjusted HR (95% CI)a  Main analysis                DMARDs  2013  18  1.28 (0.76, 2.02)  1.00 (reference)  1.00 (reference)      TNFIs  482  11  2.72 (1.36, 4.86)  2.27 (1.07, 4.80)  2.65 (1.06, 6.65)  Stratified analyses                Male                    DMARDs  592  3  0.74 (0.15, 2.17)  1.00 (reference)  1.00 (reference)          TNFIs  179  3  2.05 (0.42, 5.99)  2.91 (0.59, 14.50)  1.96 (0.32, 12.08)      Female                    DMARDs  1421  15  1.50 (0.84, 2.47)  1.00 (reference)  1.00 (reference)          TNFIs  303  8  3.10 (1.34, 6.10)  2.20 (0.93, 5.19)  2.92 (1.02, 8.32)  Analysis  n  No. of serious infections  Crude rate of infection, events/100 person-years (95% CI)  Crude HR (95% CI)  Adjusted HR (95% CI)a  Main analysis                DMARDs  2013  18  1.28 (0.76, 2.02)  1.00 (reference)  1.00 (reference)      TNFIs  482  11  2.72 (1.36, 4.86)  2.27 (1.07, 4.80)  2.65 (1.06, 6.65)  Stratified analyses                Male                    DMARDs  592  3  0.74 (0.15, 2.17)  1.00 (reference)  1.00 (reference)          TNFIs  179  3  2.05 (0.42, 5.99)  2.91 (0.59, 14.50)  1.96 (0.32, 12.08)      Female                    DMARDs  1421  15  1.50 (0.84, 2.47)  1.00 (reference)  1.00 (reference)          TNFIs  303  8  3.10 (1.34, 6.10)  2.20 (0.93, 5.19)  2.92 (1.02, 8.32)  a Adjusted for high-dimensional propensity score (tertile) in the model. HR: hazard ratio; TNFI: TNF inhibitor. Fig. 2 View largeDownload slide Three-year cumulative hazards of serious infections in children with JIA Fig. 2 View largeDownload slide Three-year cumulative hazards of serious infections in children with JIA Compared with DMARDs, use of TNFIs was associated with an increased risk of serious bacterial infection (HR = 2.74, 95% CI: 1.14, 6.56) after adjusting for the tertile of the hdPS (Table 2). In the gender-stratified analysis, an increased risk was observed (HR = 2.92, 95% CI: 1.02, 8.32) in female patients who initiated TNFIs. The point estimate was also elevated in the males (HR = 1.96, 95% CI: 0.32, 12.08) but was not statistically significant. Table 3 shows specific sites of infection in the TNFI and DMARD users. For both the TNFI and DMARD groups, infections were most commonly observed in the respiratory tract (36.4 and 33.3%), followed by the digestive system (27.3 and 22.2%), other organ systems (which included blood and device-related infections; 18.2 and 22.2%), skin and skin structure (9.1 and 16.7%) and genitourinary system (9.1 and 5.6%). Table 3 Site of infections among children with serious infection requiring hospitalizationa Site of infection  DMARDs  TNFIs  n (%)  n (%)  Number of patients with serious infection  18 (100.0)  11 (100.0)  Respiratory tract system  6 (33.3)  4 (36.4)      Upper respiratory tract infection  4 (22.2)  2 (18.2)      Pneumonia  2 (11.1)  2 (18.2)  Digestive system  4 (22.2)  3 (27.3)      Abdominal abscess  2 (11.1)  1 (9.1)      Cholecystitis  1 (5.6)  1 (9.1)      Gastroenteritis  1 (5.6)  1 (9.1)  Others  4 (22.2)  2 (18.2)      Bacteraemia/septicaemia  4 (22.2)  1 (9.1)      Device-associated infections  0 (0.0)  1 (9.1)  Skin and skin structure  3 (16.7)  1 (9.1)      Cellulitis  2 (11.1)  0 (0.0)      Necrotizing fasciitis  0 (0.0)  1 (9.1)      Septic arthritis  1 (5.6)  0 (0.0)  Genitourinary system  1 (5.6)  1 (9.1)      Pyelonephritis/urinary tract infection  1 (5.6)  1 (9.1)  Site of infection  DMARDs  TNFIs  n (%)  n (%)  Number of patients with serious infection  18 (100.0)  11 (100.0)  Respiratory tract system  6 (33.3)  4 (36.4)      Upper respiratory tract infection  4 (22.2)  2 (18.2)      Pneumonia  2 (11.1)  2 (18.2)  Digestive system  4 (22.2)  3 (27.3)      Abdominal abscess  2 (11.1)  1 (9.1)      Cholecystitis  1 (5.6)  1 (9.1)      Gastroenteritis  1 (5.6)  1 (9.1)  Others  4 (22.2)  2 (18.2)      Bacteraemia/septicaemia  4 (22.2)  1 (9.1)      Device-associated infections  0 (0.0)  1 (9.1)  Skin and skin structure  3 (16.7)  1 (9.1)      Cellulitis  2 (11.1)  0 (0.0)      Necrotizing fasciitis  0 (0.0)  1 (9.1)      Septic arthritis  1 (5.6)  0 (0.0)  Genitourinary system  1 (5.6)  1 (9.1)      Pyelonephritis/urinary tract infection  1 (5.6)  1 (9.1)  a Patients could have diagnoses of infections at more than one site at their event hospitalization, and the major site of infection was used based on clinical expert opinion. In the sensitivity analyses (Fig. 3), the risk estimate became non-significant when we restricted the study cohort to 2 to <16 years of age (HR = 2.53, 95% CI: 0.98, 6.55) and excluded patients with other rheumatoid diseases and IBD (HR = 1.15, 95% CI: 0.28, 4.17). However, we observed a greater infection risk for TNFIs when the gap for defining treatment discontinuation was reduced to 31 days (HR = 3.61, 95% CI: 1.32, 9.87) and when HCQ and SSZ users were excluded from the DMARD group (HR = 3.15, 95% CI: 1.14, 8.67). Fig. 3 View largeDownload slide Sensitivity analyses of the risk of serious infection associated with TNF inhibitors compared with DMARDs * indicates other rheumatic diseases including SLE, SS and psoriasis. ** indicates p value <0.05. DC: discontinuation; hdPS: high-dimensional propensity score. Filled circle represented adjusted HR of serious infection associated with TNFIs and the range represented 95% CI. Fig. 3 View largeDownload slide Sensitivity analyses of the risk of serious infection associated with TNF inhibitors compared with DMARDs * indicates other rheumatic diseases including SLE, SS and psoriasis. ** indicates p value <0.05. DC: discontinuation; hdPS: high-dimensional propensity score. Filled circle represented adjusted HR of serious infection associated with TNFIs and the range represented 95% CI. Discussion In a population of commercially insured children with JIA, we found that new use of TNFIs was associated with a 2.7-fold increase in risk of serious bacterial infection compared with new use of DMARDs. This increased risk estimate is consistent with the findings of some RA studies in adults [15, 16]. However, previous studies examining the association between TNFIs and infection in children with JIA had conflicting results. For example, Beukelman et al. [31, 32] analysed US national Medicaid data (2000–05) and found no difference between TNFIs and MTX in the rate of hospitalization for bacterial infections. Likewise, an analysis performed by Davies et al. [33] using data from the British Society for Paediatric and Adolescent Rheumatology Etanercept Cohort Study revealed no increased risk for serious infections requiring hospitalization and/or i.v. antibiotic use (adjusted HR 1.36, 95% CI: 0.60, 3.07) associated with etanercept compared with MTX. In contrast, Klotsche et al. [34] examined a set of German registries (2005–11) and found that the risk of infection leading to hospitalization was higher (relative risk 2.12, 95% CI: 1.08, 4.17) in patients receiving etanercept compared with MTX. Unlike previous studies, we adopted a new-user design for exposure groups. A recent study also examined new use of TNFIs and MTX in children with JIA from the Medicaid database [32]; however, in their study the use of MTX was considered a time-varying covariate, whereas in our study the effects of MTX were excluded from the TNFI group. Although TNFIs are usually recommended for use in combination with DMARDs, studies have shown that TNFIs are increasingly used early in the JIA disease course [35, 36] and as monotherapy [9, 36]. This new treatment paradigm facilitated our comparison of the two groups, as both consisted of new initiators of therapy. Notably, the black box warning for TNFIs stated that most infections developed when the drugs were used in combination with other immunosuppressants, such as DMARDs or CSs [37]. However, our study provides evidence that TNFI monotherapy is also associated with an increased risk of infection while controlling for CS use and other confounders. In addition, we further confirm the risk of infection associated with TNFI use in children with JIA. In our study, the rates of infection in the TNFI group (2.27/100 person-years) and DMARD group (1.28/100 person-years) were consistent with those observed in other studies of JIA [33, 38, 39]. The likely mechanism by which TNFIs increase the risk of infection is related to the role of TNF in the immune response [40]. Inhibiting the action of TNF prevents the normal immune response, increasing the risk for infections. It is possible that infections are not detected until they become severe because TNFIs impede the usual clinical presentation of high fever and elevated CRP. In addition, use of TNFIs may mimic a primary immunodeficiency that is considered a risk factor for mAb-related infections [41]. The time to infection (median of ∼90 days) was short in our study and was similar to findings of previous studies. For example, in the analysis by Davies et al. [33], 44% (24 of 54) of patients developed a serious infection requiring hospitalization within 6 months of etanercept initiation in children with JIA. A study of adults with RA also reported that the adjusted incidence rate ratio was 4.6 (95% CI: 1.8, 11.9) for TNFIs compared with DMARDs when restricting the follow-up time to the first 90 days [18]. These findings are consistent with the pharmacodynamic properties of TNFIs. The time to therapeutic effect of TNFIs varies by individual patient but is typically rapid; symptom improvement can be seen after two or three doses, and additional improvements over 3–6 months have been reported. Given these findings, health-care professionals should be vigilant in monitoring for symptoms of infection, especially in the first 3–6 months of treatment. Although our results suggest increased risk for infections associated with TNFIs compared with DMARDs, clinicians and patients should consider this risk in light of the benefits of TNFIs. Specifically, TNFIs are highly effective drugs that have been shown to improve symptoms, physical functioning, radiographic progression and quality of life [1–3]. In order to balance the risks and benefits associated with TNFIs, both the US FDA and the European Medicines Agency have developed risk-mitigation strategies for these biologics, but guidance on risk management for children is lacking [42, 43]. Our findings provide evidence that these agencies could use to adapt risk-management plans for children undergoing TNFI treatment. Such plans could incorporate appropriate screening, monitoring and even withholding of treatment to mitigate the potential harm of TNFIs to children with JIA. The strengths of our study include its use of relatively recent data (2009–13), use of a new-user design and use of an hdPS approach. However, our findings should be interpreted with consideration of the limitations. In particular, our results are subject to limitations common to studies using administrative claims databases, including potential misclassification of outcomes and exposure. However, we minimized this potential issue by using previously validated algorithms to identify infections and JIA [26, 27]. We also used comprehensive pharmacy records and intervention procedure codes to examine exposure, although we could not confirm that the patients administered the medications as directed. A dose–response relationship has been documented between use of CSs and higher risk of infection [44]. However, lack of information about children’s weight and prescribers’ tapering instructions resulted in considerable uncertainty in estimating the effect of the CS dose. In addition, because administrative data are not designed for research purposes, clinical data for health status in JIA (e.g. disease activity score and JIA subtype) were not available for analysis. Confounding from these and other unmeasured variables cannot be completely ruled out. Specifically, an important unmeasured confounder in our study is the relationship between JIA itself and infection. JIA or RA was found to be associated with an increased risk of infection [31, 45, 46], possibly because of T cell circulation and impaired thymic function [46, 47]. Our findings might have overestimated the TNFI–infection relationship if JIA severity was higher in the TNFI group. TNFIs are indicated for moderately to severely active polyarticular JIA, and thus children who receive TNFIs may have more severe JIA than those who receive DMARDs alone. Recognizing this possibility, we applied an hdPS approach using a large number of proxy indicators that indirectly reflected disease severity to account for the potential channelling bias. Several simulation studies have demonstrated the effectiveness of this method [28, 29]. Nonetheless, residual confounding of the JIA–infection relationship might remain. Another potential limitation of our study is insufficient statistical power. Children with JIA constitute a smaller population than adults with RA. Although the rate of TNFI-related infection in our study was not extremely low, it is still challenging for any study to obtain a large enough sample size of children with JIA to examine safety outcomes adequately, especially in subgroup analyses. As a result, we could not perform analyses to evaluate the infection risk for individual TNFI agents and for combination therapy using TNFIs and CSs, nor could we conduct analyses using only the primary diagnosis in the inpatient claims for identifying patients with severe infection. In our analysis excluding patients with other rheumatic diseases and IBD, the risk of infection reduced toward the null. On the one hand, it is possible that IBD might confound the relationship between TNFIs and serious infection, because IBD patients were more likely to receive TNFI monotherapy and develop infections. On the other hand, the small number of observed infections yielded a wide 95% CI (0.28, 4.17), and we could not draw a conclusion with insufficient power in this analysis. Future studies are needed to examine the confounding effect of IBD and confirm our findings. To facilitate studies of drug safety for children, it would be beneficial to incorporate relevant data into a national- or international-scale surveillance system. As one example, the European registry called Pharmachild is a pharmacovigilance project that aims to observe long-term adverse events associated with use of DMARDs and biologics in children with JIA across 50 participating countries [48]. In the USA, the Childhood Arthritis and Rheumatology Research Alliance has initiated a prospective observational registry enrolling patients with paediatric rheumatic diseases across North America for research purposes, including research on long-term safety of medications [49]. Also, the FDA Sentinel Initiative is a national electronic surveillance system designed to monitor proactively and examine the safety of medications and biologics [50]. In future studies, use of data from such registries and systems would support a more thorough examination of potentially serious adverse events related to TNFIs. Finally, the FDA-required black box warning might have impacted the observed association between TNFIs and infection. For example, physicians might have checked for infections more often in TNFI users than in DMARD users. In addition, the warning might have impacted the decision to admit to the hospital a patient with suspected infection differently depending on whether they were on TNFIs or DMARDs. Therefore, a differential detection or management bias might exist between the two exposure groups, which might have resulted in overestimation of relative risk in the TNFI group. In contrast, the FDA warning might have caused prescribers to avoid use of TNFIs for children susceptible to infection, and these children might have been channelled to DMARD treatment. In this case, our findings might have underestimated the relative risk in the TNFI group. However, we attempted to control for this type of confounding by applying contemporary analytical methods that included the hdPS modelling. In summary, our study demonstrated a higher risk of serious bacterial infection associated with use of TNFIs compared with DMARDs in children with JIA. Our analysis supports the FDA warning about TNFI-associated infection in children with JIA and also provides a comparison between DMARD and TNFI monotherapy. Future studies incorporating a larger cohort of children with JIA would help to confirm our findings and further characterize the risk of infection across individual TNFI agents. In the meantime, clinicians and patients need to balance the benefits of these highly effective drugs against the risk for infection they pose. Acknowledgements We gratefully acknowledge the editorial support of Mr Jon Mann, Academic Counsellor, University of Illinois at Chicago, during the preparation of this manuscript. The views expressed in this article are those of the authors and do not necessarily represent the views of the Department of Veterans Affairs or Health Services Research and Development Service. Funding: No specific funding was received from any bodies in the public, commercial or not-for-profit sectors to carry out the work described in this manuscript. Disclosure statement: G.T.S. has received payment for consulting or speaking from AbbVie Inc., Astellas Pharma US Inc. and Baxter International Inc., in the past 3 years. All other authors have declared no conflicts of interest. Supplementary data Supplementary data are available at Rheumatology Online. References 1 Lovell DJ, Giannini EH, Reiff A et al.  ; Pediatric Rheumatology Collaborative Study Group. Etanercept in children with polyarticular juvenile rheumatoid arthritis. N Engl J Med  2000; 342: 763– 9. Google Scholar CrossRef Search ADS PubMed  2 Lovell DJ, Ruperto N, Goodman S et al.   Adalimumab with or without methotrexate in juvenile rheumatoid arthritis. N Engl J Med  2008; 359: 810– 20. Google Scholar CrossRef Search ADS PubMed  3 Keystone EC, Kavanaugh AF, Sharp JT et al.   Radiographic, clinical, and functional outcomes of treatment with adalimumab (a human anti-tumor necrosis factor monoclonal antibody) in patients with active rheumatoid arthritis receiving concomitant methotrexate therapy: a randomized, placebo-controlled, 52-week trial. Arthritis Rheum  2004; 50: 1400– 11. Google Scholar CrossRef Search ADS PubMed  4 Beukelman T, Patkar NM, Saag KG et al.   2011 American College of Rheumatology recommendations for the treatment of juvenile idiopathic arthritis: initiation and safety monitoring of therapeutic agents for the treatment of arthritis and systemic features. Arthritis Care Res  2011; 63: 465– 82. Google Scholar CrossRef Search ADS   5 Singh JA, Saag KG, Bridges SLJr et al.   2015 American College of Rheumatology Guideline for the treatment of rheumatoid arthritis. Arthritis Rheumatol  2016; 68: 1– 26. Google Scholar PubMed  6 Breedveld FC, Emery P, Keystone E et al.   Infliximab in active early rheumatoid arthritis. Ann Rheum Dis  2004; 63: 149– 55. Google Scholar CrossRef Search ADS PubMed  7 St Clair EW, van der Heijde DM, Smolen JS et al.   Combination of infliximab and methotrexate therapy for early rheumatoid arthritis: a randomized, controlled trial. Arthritis Rheum  2004; 50: 3432– 43. Google Scholar CrossRef Search ADS PubMed  8 Lee SJ, Chang H, Yazici Y et al.   Utilization trends of tumor necrosis factor inhibitors among patients with rheumatoid arthritis in a United States observational cohort study. J Rheumatol  2009; 36: 1611– 7. Google Scholar CrossRef Search ADS PubMed  9 Mannion ML, Xie F, Curtis JR, Beukelman T. Recent trends in medication usage for the treatment of juvenile idiopathic arthritis and the influence of tumor necrosis factor inhibitors. J Rheumatol  2014; 41: 2078– 84. Google Scholar CrossRef Search ADS PubMed  10 Schneeweiss S, Setoguchi S, Weinblatt ME et al.   Anti-tumor necrosis factor alpha therapy and the risk of serious bacterial infections in elderly patients with rheumatoid arthritis. Arthritis Rheum  2007; 56: 1754– 64. Google Scholar CrossRef Search ADS PubMed  11 Grijalva CG, Chen L, Delzell E et al.   Initiation of tumor necrosis factor-α antagonists and the risk of hospitalization for infection in patients with autoimmune diseases. JAMA  2011; 306: 2331– 9. Google Scholar CrossRef Search ADS PubMed  12 Bongartz T, Sutton AJ, Sweeting MJ et al.   Anti-TNF antibody therapy in rheumatoid arthritis and the risk of serious infections and malignancies: systematic review and meta-analysis of rare harmful effects in randomized controlled trials. JAMA  2006; 295: 2275– 85. Google Scholar CrossRef Search ADS PubMed  13 Bernatsky S, Habel Y, Rahme E. Observational studies of infections in rheumatoid arthritis: a metaanalysis of tumor necrosis factor antagonists. J Rheumatol  2010; 37: 928– 31. Google Scholar CrossRef Search ADS PubMed  14 Alonso-Ruiz A, Pijoan JI, Ansuategui E et al.   Tumor necrosis factor alpha drugs in rheumatoid arthritis: systematic review and metaanalysis of efficacy and safety. BMC Musculoskelet Disord  2008; 9: 52. Google Scholar CrossRef Search ADS PubMed  15 Curtis JR, Patkar N, Xie A et al.   Risk of serious bacterial infections among rheumatoid arthritis patients exposed to tumor necrosis factor α antagonists. Arthritis Rheum  2007; 56: 1125– 33. Google Scholar CrossRef Search ADS PubMed  16 Lane MA, McDonald JR, Zeringue AL et al.   TNF-α antagonist use and risk of hospitalization for infection in a national cohort of veterans with rheumatoid arthritis. Medicine  2011; 90: 139– 45. Google Scholar CrossRef Search ADS PubMed  17 Listing J, Strangfeld A, Kary S et al.   Infections in patients with rheumatoid arthritis treated with biologic agents. Arthritis Rheum  2005; 52: 3403– 12. Google Scholar CrossRef Search ADS PubMed  18 Dixon WG, Symmons DPM, Lunt M et al.   Serious infection following anti-tumor necrosis factor α therapy in patients with rheumatoid arthritis: lessons from interpreting data from observational studies. Arthritis Rheum  2007; 56: 2896– 904. Google Scholar CrossRef Search ADS PubMed  19 Salmon-Ceron D, Tubach F, Lortholary O et al.   Drug-specific risk of non-tuberculosis opportunistic infections in patients receiving anti-TNF therapy reported to the 3-year prospective French RATIO registry. Ann Rheum Dis  2011; 70: 616– 23. Google Scholar CrossRef Search ADS PubMed  20 Information for Healthcare Professionals: Cimzia (certolizumab pegol), Enbrel (etanercept), Humira (adalimumab), and Remicade (infliximab). http://www.fda.gov/Drugs/DrugSafety/PostmarketDrugSafetyInformationforPatientsandProviders/ucm124185.htm. (Accessed on 12 November 12, 2016, date last accessed). 21 Kaushal R, Bates DW, Landrigan C et al.   Medication errors and adverse drug events in pediatric inpatients. JAMA  2001; 285: 2114– 20. Google Scholar CrossRef Search ADS PubMed  22 World Health Organization. Promoting Safety of Medicinces For Children. 2007. http://www.who.int/medicines/publications/essentialmedicines/Promotion_safe_med_childrens.pdf (12 November 2016, date last accessed). 23 Ruperto N, Lovell DJ, Cuttica R et al.   A randomized, placebo-controlled trial of infliximab plus methotrexate for the treatment of polyarticular-course juvenile rheumatoid arthritis. Arthritis Rheum  2007; 56: 3096– 106. Google Scholar CrossRef Search ADS PubMed  24 Horneff G, De Bock F, Foeldvari I et al.   Safety and efficacy of combination of etanercept and methotrexate compared to treatment with etanercept only in patients with juvenile idiopathic arthritis (JIA): preliminary data from the German JIA Registry. Ann Rheum Dis  2009; 68: 519– 25. Google Scholar CrossRef Search ADS PubMed  25 Danielson E. White Paper. Health Research Data for the Real World: The MarketScan® Databases. 2011. http://truvenhealth.com/portals/0/assets/PH_11238_0612_TEMP_MarketScan_WP_FINAL.pdf (23 March 2016, date last accessed). 26 Harrold LR, Salman C, Shoor S et al.   Incidence and prevalence of juvenile idiopathic arthritis among children in a managed care population, 1996–2009. J Rheumatol  2013; 40: 1218– 25. Google Scholar CrossRef Search ADS PubMed  27 Patkar NM, Curtis JR, Teng GG et al.   Administrative codes combined with medical records based criteria accurately identified bacterial infections among rheumatoid arthritis patients. J Clin Epidemiol  2009; 62: 321– 7, 327 e1–7. Google Scholar CrossRef Search ADS PubMed  28 Schneeweiss S, Rassen JA, Glynn RJ et al.   High-dimensional propensity score adjustment in studies of treatment effects using health care claims data. Epidemiology  2009; 20: 512– 22. Google Scholar CrossRef Search ADS PubMed  29 Rassen JA, Glynn RJ, Brookhart MA, Schneeweiss S. Covariate selection in high-dimensional propensity score analyses of treatment effects in small samples. Am J Epidemiol  2011; 173: 1404– 13. Google Scholar CrossRef Search ADS PubMed  30 Pharmacodynamics and Pharmacokinetics section. Drug information. UpToDate. http://www.uptodate.com (20 July 2016, date last accessed). 31 Beukelman T, Xie F, Chen L et al.   Rates of hospitalized bacterial infection associated with juvenile idiopathic arthritis and its treatment. Arthritis Rheum  2012; 64: 2773– 80. Google Scholar CrossRef Search ADS PubMed  32 Beukelman T, Xie F, Baddley JW et al.   The risk of hospitalized infection following initiation of biologic agents versus methotrexate in the treatment of juvenile idiopathic arthritis. Arthritis Res Ther  2016; 18: 210. Google Scholar CrossRef Search ADS PubMed  33 Davies R, Southwood TR, Kearsley-Fleet L et al.   Medically significant infections are increased in patients with juvenile idiopathic arthritis treated with etanercept: results from the British Society for Paediatric and Adolescent Rheumatology Etanercept Cohort Study. Arthritis Rheumatol  2015; 67: 2487– 94. Google Scholar CrossRef Search ADS PubMed  34 Klotsche J, Niewerth M, Haas JP et al.   Long-term safety of etanercept and adalimumab compared to methotrexate in patients with juvenile idiopathic arthritis (JIA). Ann Rheum Dis  2016; 75: 855– 61. Google Scholar CrossRef Search ADS PubMed  35 Otten MH, Anink J, Prince FH et al.   Trends in prescription of biological agents and outcomes of juvenile idiopathic arthritis: results of the Dutch national Arthritis and Biologics in Children Register. Ann Rheum Dis  2015; 74: 1379– 86. Google Scholar CrossRef Search ADS PubMed  36 Lee WJ, Briars L, Lee TA et al.   Use of tumor necrosis factor-alpha inhibitors in children and young adults with juvenile idiopathic arthritis or rheumatoid arthritis. Pharmacotherapy  2016; 36: 1201– 9. Google Scholar CrossRef Search ADS PubMed  37 Product Information: ENBREL (etanercept) for subcutaneous injection. Amgen, Inc., Thousand Oaks, CA. 2016. http://www.accessdata.fda.gov/drugsatfda_docs/label/2015/ 103795s5548lbl.pdf (8 March 2017, date last accessed). 38 Lovell DJ, Reiff A, Ilowite NT et al.   Safety and efficacy of up to eight years of continuous etanercept therapy in patients with juvenile rheumatoid arthritis. Arthritis Rheum  2008; 58: 1496– 504. Google Scholar CrossRef Search ADS PubMed  39 Tarkiainen M, Tynjälä P, Vähäsalo P, Lahdenne P. Occurrence of adverse events in patients with JIA receiving biologic agents: long-term follow-up in a real-life setting. Rheumatology  2015; 54: 1170– 6. Google Scholar CrossRef Search ADS PubMed  40 Wajant H, Pfizenmaier K, Scheurich P. Tumor necrosis factor signaling. Cell Death Differ  2003; 10: 45– 65. Google Scholar CrossRef Search ADS PubMed  41 Maródi L, Casanova JL. Primary immunodeficiencies may reveal potential infectious diseases associated with immune-targeting mAb treatments. J Allergy Clin Immunol  2010; 126: 910– 7. Google Scholar CrossRef Search ADS PubMed  42 European Medicines Agency. http://www.ema.europa.eu/ema (5 July 2016, date last accessed). 43 Approved Risk Evaluation and Mitigation Strategies (REMS). US Food and Drug Administration. http://www.accessdata.fda.gov/scripts/cder/rems (5 July 2016, date last accessed). 44 Lovell DJ, Zaoutis TE, Sullivan K. Immunosuppressants, infection, and inflammation. Clin Immunol  2004; 113: 137– 9. Google Scholar CrossRef Search ADS PubMed  45 Smitten AL, Choi HK, Hochberg MC et al.   The risk of hospitalized infection in patients with rheumatoid arthritis. J Rheumatol  2008; 35: 387– 93. Google Scholar PubMed  46 Blumentals WA, Arreglado A, Napalkov P, Toovey S. Rheumatoid arthritis and the incidence of influenza and influenza-related complications: a retrospective cohort study. BMC Musculoskelet Disord  2012; 13: 158. Google Scholar CrossRef Search ADS PubMed  47 Koetz K, Bryl E, Spickschen K et al.   T cell homeostasis in patients with rheumatoid arthritis. Proc Natl Acad Sci USA  2000; 97: 9203– 8. Google Scholar CrossRef Search ADS PubMed  48 Swart J, Consolaro A, Horneff G et al.   Pharmacovigilance in juvenile idiopathic arthritis patients (Pharmachild) treated with biologic agents and/or methotrexate. Consolidated baseline characteristics from Pharmachild and other national registries. Pediatr Rheumatol Online J  2014; 12: P7. Google Scholar CrossRef Search ADS   49 The Childhood Arthritis and Rheumatology Research Alliance (CARRA) registry. https://carragroup.org/carra-registry-updates/ (2 November 2016, date last accessed). 50 FDA's Sentinel Initiative. http://www.fda.gov/Safety/FDAsSentinelInitiative/ucm2007250.htm (20 June 2016, date last accessed). © The Author 2017. Published by Oxford University Press on behalf of the British Society for Rheumatology. All rights reserved. For Permissions, please email: journals.permissions@oup.com

Journal

RheumatologyOxford University Press

Published: Feb 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off